compositional variations of .tt?f;yil1ff?;11,[ degree of ... · american mineralogist, volume 62,...

American Mineralogist, Volume 62, pages 843-865, 1977

Compositional variations of ."Tt?f;y"il1ff?;11,[ degree of melting of peridotite

BldnN O. MysnN eNo Iruo KusnlRol

Geophysical Laboratory, Carnegie Institution of WashingtonWashington, D. C.20008

Abstract

A new technique using radioactive tracers has been developed for accurate determination ofthe degree of melting of rocks. Trace amounts of a compound (CO, is used here) or an elementthat preferentially enters the melt phase are determined. A glass of the starting material is usedas standard. The proportion of partial melt equals the ratio of the amount of tracer in thestandard and the unknown.

Using this technique, two garnet peridotite nodules, samples l6l I (from a Lesotho kimber-l ite) and 665,{L-l (from Hawaiian nephelinite tuff), were subjected to temperatures from theirsolidi to 1750'C aL 20 and,35 kbar pressure. Both peridotite nodules show similar meltingbehavior. The degree of melting was plotted against temperature, and such melting curvesshow a distinct change of slope wherever one or more phases disappear. At 20 kbar, olivine(Ol), cl inopyroxene (Cpx), orthopyroxene (Opx), and spinel (Sp) coexist with tholeiit ic l iquiddu r i ng the in i t i a l 25pe rcen tme l t i ngo f samp le l 6 l l and60pe rcen tme l t i ngo f 6654 'L - l .Thetemperature interval of tholeiite melt * 2 pyroxenes * olivine is less than 40'C in both cases.

At 35 kbar, quality chemical analysis of the samples became impossible because of quench-ing problems. Nodule l6l1 begins to melt at about 1625'-1630'C, apparently producingalkali picrit ic l iquid coexisting with garnet, two pyroxenes, and olivine unti l about 25 percentmelting has occurred and the temperature has risen to about 1660'C, where garnet disappears.

The flat slope of the melting curves in the lowest-temperature melting interval and thenearly constant bulk composition of the l iquid suggest that the tholeiit ic l iquid in this meltingrange may be derived from a peridotit ic parent under near isobaric, invariant conditions.

ln the lowest-temperature melting interval at 20 kbar two reaction relations between partialmelt and crystall ine residue are suggested: for 665,{L-l (lowest temperature) the reaction is Sp* Opx * Cpx (Di"") = Ol * Liq; for the higher temperature melting of l6l I the reaction is Ol* Cpx (pigeonite) = Opx * Liq. The observations suggest that phase equil ibria in simplesystems such as MgSiOr-CaSiOB-Al2OB and MgrSiOo-CaMgSizOe-SiO, closely resemblethose of natural peridotite in the upper mantle.

Introduction

Accurate determination of the degree of melting ofrocks as a function of intensive variables (pressure,temperature, and activity of volati les) is an importantaid to understanding the processes of partral meltingand magma genesis. Data on melting behavior ofrocks are also useful for understanding major andtrace element behavior in l iquids and crystals duringpartial melting and fractional crystall ization.

Ever since Bowen (1928) suggested a peridotite

1 Present address: Geological Inst i tute, Univers i ty of Tokyo,Hongo , Tokyo I 13 , Japan .

source of basalt, genesis of magma by partial meltingof peridotitic upper mantle has frequently been dis-cussed either in terms of phase equilibria in appropri-ate model systems [see Yoder (1976) for discussionand earlier references]. or simply in terms of varyingdegrees of melting of peridotite compositions (e.9.,Green and Ringwood, 1967; Green, 1970). In thisstudy the primary objective was to evaluate thesepossibilities under volatile-free conditions, althougha few experiments involving HrO are included. Vol-atiles are important in the upper mantle (for review ofdata see lrving and Wyllie, 1975; Boettcher et al.,1975; Anderson, 1975). Experimentat ion in naturaland model systems has demonstrated the important

843

MYSEN AND KUSHIRO: COEXISTING PHASES OF PERIDOTITE

effects of such components in the partial melting ofperidotitic upper mantle (e.g., Kushiro et al., 1968;Kushiro, 1969, 1972; Mysen and Boettcher, l975a,b:Eggler, 1974,1975). As a first attempt to study quan-titatively the phase equilibria of partial melting ofcomplex natural peridotite, the added complexity ofliquid-crystal-vapor equilibria has not been consid-ered.

Experimental technique

Starting materials were two garnet peridotite nod-ules of the compositions given in Table l. Nodulel6l I is a sheared garnet peridotite from a L€sothokimberl i te descr ibed by Nixon and Boyd (1973). Themajor element data of Nixon and Boyd (1973) alongwith the rare-earth element data of Shimizu (1974)indicate that this nodule represents the least depletedupper mantle among kimberlite nodules from SouthAfrica. Preliminary experimental data on sample16l l were reported by Kushiro (1973). The othernodule, 6654L-1, is a garnet per idot i te from a Ha-waiian nephelinite tuff of the Honolulu series firstdescribed by Jackson and Wright (1970). Shaw andJackson (1973) suggested that nodule 665,{L-l is themost representative of undepleted upper mantle be-neath the Hawai ian Is lands. Nodule 66SAL-1 haspreviously been studied at high pressure in the pres-ence of HrO by Mysen and Boettcher (1975a,b).

Both samples were crushed to less than 5 pm grainsize and fired at I150'C withl(Or) controlled at thelevel of the QFM buffer before use. After crushingthe original nodule to about 70 1rm grain size, a 0.59

Table l , Composi t ions of star t ing mater ia ls

66sAL- 1 i

split was crushed under acetone for about 90 minutesto reduce the grain size to (5 pm. The heat treatmentdid not significantly alter the garnet peridotite miner-alogy except for a slight tendency toward breakdownof the garnet crystals along their edges. Sample l6l Iwas also studied with an amount of HrO equal to thatbound in hydroxylated minerals (1.9 weight percentHzO; Nixon and Boyd, 1973). The hydroxylated min-erals are fine-grained phlogopite and serpentine(F. R. Boyd, personal communicat ion, 1976).

The degree of melting of a rock can be determinedby adding a trace amount of a compound (e.g., COr)that enters only the liquid phase. The partitioncoefficient of COz between liquid and crystals exceeds1000 (Mysen et al., 1976), and this compound wastherefore selected for this purpose. To this end, ap-proximately 50 ppm COz (as BaCOs) with a knownamount of radiogenic loC was added to each sample.The amount of melt in the sample was determined bybeta-track counting of unknown and standard, util iz-ing the 'aC spike as described by Mysen and Seitz(1975). The standard was a glass of the starting mate-rial with 'oC-spiked CO, added. The percentage ofmelt in the unknown is then simply

percentage melt = [(carbon in all glassstandard)/(carbon in glass of partially

melted unknown)l X 100 ( l )

More than 2 weight percent CO, dissolves in basal-tic melts at pressures above 20 kbar (Mysen el a/.,1975). Therefore, with 50 ppm COz added to thesample, no COz vapor will be present in equilibriumwith COr-saturated partial melt when the amount ofmelt is larger than 0.25 percent. The absence of vaporis a requirement for accurate determination of thepercentage of melt in a charge when using laC-spiked

COz as the analyt ical tool . In sample 16l l + 1.9weight percent HrO, this requirement presented aproblem inasmuch as HzO-rich vapor may be presentin the 5-10 percent melting range. No modificationsof the technique were made to overcome this prob-lem, however, because of the relatively slight impor-tance assigned to this part ofthe system in the presentexperiments. In general, the problem can be solved bychanging to another radioactive isotope that does notdissolve in vapor and does not enter crystalline sili-cates in significant amounts (e.9., tungsten-185).

Carbon dioxide is formed in the experiment bybreakdown of finely-disseminated BaCOs duringheating prior to melting of the sample. Mysen andSeitz (1975) have shown that a run durat ion of 5minutes is necessary to attain equilibrium CO, con-

s lo2T i02Al203Fe2O3FeOMnOMcoCaONa2OK20H2o-H20-P205Cr203Nlo

Tota ls

43 .700 . 2 5) 1 \

1 , 3 88 . 8 10 , 1 3

a 1 ) )

3 . 2 60 . 3 30 . 140 . 0 5r , 9 4

0 . 2 8n . d .

700.24

44 .82o . 5 26 . Z L

2 . 0 77 . 9 10 . 1 9

26,53e 1 ,

0 . 8 90 . 0 30 . 1 10 . I 50 .040 . 200 . 20

99 .99

*Ana lysLs f rm N lxon and Boyd (1973) ,tAna lys is f rm Mysen and Boet tcher (1975a) .Both ana lyses o f un f i red eanp les . The ana lys is o f 1611

cor responds to 1611 + 1 .9 we igh t percent H2O in the tex t .

MYSEN AND KUSHIRO: COEXISTING PHASES OF PERIDOTITE 845

tents of a si l icate l iquid of albi te composit ion at1450'C and 20 kbar. Becarrse the present experimentswere at similar temperatures and the run durationsfar exceeded 5 minutes (Table 2), equilibrium was inal l l ikel ihood attained.

The uncertainties in the data include the countingstatistics, heterogeneities of carbon in the samples,and the exposure times. As seen from the data inTable 2, the total relative uncertainty (+ lo) is gener-ally less than 5 percent.

Crystallization of quench phases (pyroxene andolivine) is sometimes a problem in chemical analysisof the phases present in the experiments. Fortunately,carbon remains in the quench minerals at least on anoptical scale (needed for beta-track counting). Thisoccasional problem therefore did not affect the re-sul ts of the determinat ion of the degree of melt ing.

All experiments were carried out in a solid-media,high-pressure apparatus similar to the design of Boydand England (1960). The experiments were done bythe piston-out technique with no correction for fric-

tion, resulting in A 1.5 kbar uncertainty. Temper-atures were measured with Pt-Pt90Rhl0 thermo-couples without pressure correction on emf output.The absolute accuracy of the temperature measure-ments is about t l0oC. However, the automatic tem-perature controllers maintain the nominal temper-atures to within 3oC, which is probably theapproximate precision in these experiments. Con-sequently, temperature effects could be observedwithin about 5oC.

The samples were contained in sealed PtruAuu con-tainers. Sealed capsules were used to avoid loss ofcarbon dioxide. Glass-sleeved furnace assemblies of0.5 inch diameter were used in all experiments tominimize access of hydrogen (from HrO releasedfrom the outer talc sleeve) to the sample area. Thisexperimental configuration was employed to reducethe iron loss to the capsule by reducing the effect ofhydrogen fugacity on the oxidation state of iron in-side the capsule. However, the PtruAuu capsules stil ldissolved sufficient iron to result in about 20 percent

Table 2 Data on exper imental runs

Run No. T : *P.( "c )

Pres s u re(kbar )

Run Dura t ion(n in . )

Phases Present Percent MeI t

S tar t lns Mater ia l 66SAL-1

4-494-57+ -+ )4-50+ - ) Y

4-844 -7 04 -624-7 4

4 -8 r

4-7 L

4 -L094 - I I U

4-LL4

+ - I I Y

4 - I294 _ I J 1

4L24 t r

409

+ L J

4 0 8

410

13 50l J O J

I37 514001500

I 4 O U

L47 5I 500

15 501600r7 50

1450L47 51500L ) Z )

1 550

I60017 0016 10L 0 a )

16 50

I O O J

I O 6 )

17 0017 50

S t a r t lng

2020202020

240190180L206 0

Ma te r l a l 1611 + 1 .9

01 ,0 1 ,0 1 ,0 1 ,0 1 ,

welqht Dercent

Opx, Cpx,Opx, Cpx,Opx, Cpx,Opx, L iqL t q

HrO

Opx, Cpx,Opx, Cpx,o p x , C p x ,Opx, L iq

L iqL tqL i q

Opx, Cpx,Opx, Cpx,opx , Cpx ,o p x , L l qopx , L tq

L i qL l qOpx, Cpx,Opx, Cpx,Opx, Cpx,

opx , Cpx ,Opx, Cpx,Opx, L iqOpx, L iq

I , 4 9 + O . O 71 1 . 5 + 0 . 6q O 7 r 1 A

8 5 + 59 7 + 2

6 . 1 + 0 . 3

5 9 + 36 0 . 3 + 2 , 71 ? 1 L a 1

L . 3 2 + 0 , 0 4L . 7 7 + 0 . 0 5

I 5 . 7 + 0 . 528 .5 + O .73 4 . 5 + 0 . 8

4 4 . L + L . 56 0 . 8 + 1 . 20 . 6 9 + 0 . 0 10 . 9 2 + o . o 2

1 3 . 6 + 0 . 3

? ? O r n q

4 3 . 6 + 1 . 0+ + . ) + L . Z

o r . + T r . o

S p , L i qs p , L i qS p , L l q

L tqL l qL l q

20202020

202020

20z0202020

2020J )

35J J

J f

J )

J f

J )

180180105120

0 1 ,0 1 ,0 I ,o I ,

0 1 ,0 1 ,O I ,

2L06 0

5

Star t lns Mater ia l 16 l I

180180180150L20

7 57 590907 0

906 03020

0 1 ,0 1 ,0 1 t

0 1 ,0 1 ,

o I t0 1 ,0 1 ,0 1 ,0 1 ,

0 1 '0 1 ,o l ,0 1 ,

L iqI-1qL l q

ca , L iqGa, l , lqG a , L t q

L tq


iron loss from the samples at 20 kbar pressure at thehighest temperatures ( )1550'C). The i ron loss wasconsiderably higher at 35 kbar. This increase athigher pressure is ascribed to the much lower vis-cosity of sil icate l iquids at 35 kbar compared with 20kbar (65 percent decrease; Kushiro et al.,1976), re.sulting in higher diffusion rates and therefore morerapid transport of iron from the charge to the cap-sule. Alternatives to simple noble-r.netal capsules,such as graphite capsules inside platinum capsules,had to be discarded because of exchange of'nC be-tween graphite and the sil icate charge. Consequently,it was decided to accept the iron loss, even thoughthat problem may produce unc€rtainties in applyingvariation of iron content during fhe experiments. Forexample, 20 percent iron loss in olivine results in - Ipercent increase of Fo contont with Foro composi-tion. Such uncertainties are mpch smaller than theobserved temperature effeits and therefore do notaffect the conclusions significantly.

All analytical data were obtained with the MACautomated electron microprobe at the GeophysicalLaboratory using l5 kV acceleration voltage.

Quenched l iqu ids (g lass) were analyzed wi th 0.01 pAbeam current to reduce loss gf alkalies by vol-ati l ization. The data show that even with such pre-cautions, 50 percent or more NarO and some K2Owere lost from the glasses. However, no further at-tempts were made to correct the alkali loss in theanalyses presented.

Results

Phase equilibria

Data on the experimental runs appear in Table 2.Degree of melting [X(melt)] is plotted against tem-perature (7) for nodules 16l l and 66SAL-I at 20kbar in Figure l. The model composition FourEn.oDi,, (by weight), used to i l lustrate melting behaviorof s imple systems in th is k ind of d iagram, is rep-resentative of typical peridotite.

The identif ication,of phase fields as presented inthe figures is based on optical data and confirmationby the electron microprobe. Whether spinel occurstogether with olivine and two pyroxenes in samplel6 l I (F ig. l ) is uncer ta in. A few gra ins ( 1 pm acrossand apparently isotropic, may be spinel, but this ob-servation could not be confirmed with the electronmicroprobe. Spinel is therefore left out of the figure.Clinopyroxenes are called Ca:rich or Ca-poor on thebasis of thei r chemical composi t ions. Kushiro (1969)and Kushiro and Yoder (1970) suggested that pi-

F ig . l . Me l t i ng cu rves o fnodu les l 6 l l and 66SAL- l a t 20 kba rpressure (anhydrous) Size of boxes includes uncerta int ies in

temperature (+ l0"C) and determinat ion of percentage of l iquid( to) . ForuEnroDi,u curve f rom Kushiro (1969).

geonite is stable to at least 20 kbar. Some, if not all,o f the Ca-poor c l inopyroxenes in l6 l1 could be p i -geonitp (see discussion below), whereas all the cli-nopyroxenes of 665AL-l are Ca-rich (diopsidic).

Starting from the solidi (Fig. I ), the melting rangesof these peridotites are divided into three distinctlydifferent f ields, olivine -t 2 pyroxenes + l iquid (+spinel ) , o l iv ine f or thopyroxene * l iqu id, and o l i -v ine * l iqu id. There is a lso a st r ik ing s imi lar i ty be-tween these patterns and that of compositionFo ruEns .D i r . i n t he t h ree -componen t sys temDi-Fo-SiOz (Kushiro, 1969). Sol id solut ions in-volving Fe, Al, Cr, and Na in the natural samples do,of course, result in slightly different curvature of themelting curves of the natural samples compared withthat of the simple system. ThE init ial melting step (Ol* 2 Px [+ Sp] + L iq) of both 16l l and 66SAL-Iclosely resembles the invariant melting behavior ofthe synthetic composition (Fig. l), suggesting thatvolati le-free melting of natural peridotite may ap-proximate an invariant character. Both meltingcurves of the natural rocks show a sharp dip close totheir solidi without change of phase assemblage. Thisfeature does not occur in the simple system. Bothnodules contain several tenths ofa percent alkalies. Itmay be that such cations (chainbreakers or depoly-mer izers [Bot t inga and Wei l l , 1972;Waff ,1975]) in-

E

20 kbo r Fo55En39D i15 - - , , ' /

.C , /

^1",< - / /,' /'-, , / \

, ' N -t6t' l

-^*.99-- zN. o . P ) - . _ o , /

!-9+ upj+_ugx:Llq/ /

r ,0 50P e r c e n t m e l t

MYSEN AND KUSHIRO: COEXISTING PHASES OF PERIDOTITE 847

duce a small amount of melting at a considerablylower temperature than in their absence. Qual-itatively, this melting behavior resembles that ofHrO-deficient rocks in the absence of hydrous miner-als (as defined by Merri l l et al., 1970). Clearly, themel t ing is not invar iant in th is region.

The solidus temperature of rock 161I was reportedas 1475'C at 20 kbar by Kushiro (1973). I f the mel t -ing curvg of l6l I in Figure I is extrapolated l inearlyfrom teinperatures immediately above the sharp dip,the sol idus temperature a lso becomes l475oC. Sim-i lar ly , the sol idus of 6654L- l becomes 1350'C at 20kba r .

Samples l6 l1 and 665AL- l are quant i ta t ive ly d i f -lerent in ternis of temperature and degree of meltingat which phase changes occur. The degree of ineltingof sample 665,4L-l increases from 1.5 to 60 percentbetween 1350' and 1375"C at 20 kbar wi thout changeof phase assemblage. In comparison, only 25 percentmelt can be derived from 1611 before clinopyroxenedisappears (F ig. l ) .

Water-free and water-undersaturated melting ofl6l I at 20 kbar is shown in Figure 2. The presence ofa smal l amount of water in l6 l l adds two technicalproblems. First, as discussed above, vapor may occuras a separate phase at the lowest degree of melting

( the par t ia l mel t o f l6 l1 + 1.9 weight percent HrO isprobably water-saturated with the degree of meltingbelow about 7 percent). Under such conditions somecarbon dioxide partit ions into the HrO-rich vapor,and the amourit of CO, left in the partial melt is lessthan that for a similar proportion of melt underwater-free corldit ions. Consequently, the 6.1 percentmelting shown in Figure 2 may be an overestimate.Second, the presence of quench minerals, formingmore readily from hydrous sil icate melt, results ingreater diff iculty in analyzing the run products of16l l + 1.9 weight percent HrO. The two mel t ingcurves obtained with nodule l6l I at 20 kbar (HzO-free and with 1.9 weight percent HrO) are quite sim-ilar, except that the hydrous curve l ies at lower tem-peratures. The temperature difference increases from25'C wi th in the o l iv ine * 2 pyroxene * l iqu id f ie ld tomore than 150'C at 60 percent melting. The differ-ence in,the degree of melting at a given temperatureinside the melting range where pyroxene is present isas much as 25 percent (absolute) because of the flatslope of the melting curves in this temperature.region.

The resul ts f rom sample 16l l a t 2O and 35 kbarpressure are shown in Figure 3. The melting intervalat 35 kbar can be subdivided into four distihct re-gions, although the highest-temperature field in Fig-

/ ! J 1 7

o

ooEo 1 6

1 8 0 0

0 0

0 0

1 5 0 0

1611; 20 k bo r

1 . 9 w t % H 2 0

1 4 0 00 1 0 2 0 3 0 L 0 5 0 6 0 7 0 8 0 9 0 1 0 0

P e r c e n t m e l t

F ig. 2. Mel t ing curves of 16l I anhydrous and in the presence of 1.9 weight percent H,O at 20 kbar. Size of boxes as in Fig


ure 3 was not determined. The stippled line and thebracketed "Ol + Liq" field are merely a suggestion,constrained only by the improbability of the 35 kbarmelting curve intersecting the 20 kbar melting curve.

The phase assemblages at 35 kbar are differentfrom those at 20 kbar in two respects. First, garnetappears on the 35 kbar solidus and remains stable toabout 25 percent melting. Apart from the sharp dipbelow about I percent melting, the interval of garnetstability is about 40oC. Second, pyroxene has a muchwider stability interval at 35 kbar than at 20 kbar (to-60 and 45 percent melting, respectively, at 35 and20 kbar; Fig. 3). Because of the steeper slopes of the20-kbar melting curve, the individual intervals in-clude about the same amount of melt (excluding, ofcourse, the interval of garnet stability, which has no20-kbar analogue). The melting intervals above thatof garnet stability involve the same minerals as thoseat 20 kbar (Fig. 3).

The ambiguity of the solidus temperature at 35kbar is the same as at 20 kbar. Extrapolation of the35 kbar melt ing curve simi lar to that of the 20 rbarmelting curve results in a solidus temperature of1625'-1630'C for 161 1. The data of Kushiro (1973),extrapolated linearly from 25 kbar, suggest l660oC asthe 35 kbar sol idus temperature.

Analytical data

Reliable chemical data were obtained on samplesheld at 20 kbar. Quenching problems became increas-

O t + 0 p x + C P xL r q

ingly cumbersome as pressure, temperature, and HzOand alkali contents of the system increased. The mostcommon quench mineral is an o l iv ine ( -43 weightpercent SiOr) with 2-4 weight percent AlrO, and CaO(submicroscopic l iquid inclusions?) and withMg/(Mg * Fe) : 0.89-0.90. Quench olivine is com-monly optically indistinguishable from stable olivine,and its identif ication is based mostly on its chemicalcomposition. The hydrous runs also resulted inquench clinopyroxene containing - 10 weight percentCaO and - 10 weight percent AlzO' and havingMg/(Me + Fe) : 0 .7-0.75. The presence of s igni f i -cant amounts of such quench clinopyroxene insample l6 l l + 1.9 weight percent HzO made i t par-ticularly diff icult to obtain reliable analyses of Ca-poor clinopyroxene from this sample. Similar quenchpyroxene precipitated from liquids of sample 665,4L-I, but because of the greater compositional differencebetween stable and quench clinopyroxene in this ma-ter ia l , re l iab le data could be obta ined. Analys is ofquenched l iqu id (g lass) f rom HrO-f ree 161I was s im-pler because the main quench mineral was olivine.The total CaO and AlrO, contents of l iquids in-creased dramatically in l iquid volumes supplyingcomponents to such quench minerals. Quality l iquidanalysis could be obtained by moving the electronbeam away from the quench mineral unti l the CaOand AlrOr contents of the quenched glass werelowered to a fixed value.

All experiments at 35 kbar underwent very exten-sive quench-mineral crystall ization, including botholivine and two pyroxenes. The problem proved suf-ficiently severe to abort attempts to obtain qualityanalytical data from such experimental run products.

Oliuine. Relevant olivine compositional trends forboth nodules as a function of degree of melting areshown in Figure 4. Their Mgl(Mg + Fe) (Fo con-tent) define the same distinct melting ranges as shownby the melting curves (Figs. I and2). The most rapidchanges of Fo content occurjust above the peridotitesolidus and in the olivine f l iquid fields, suggestingthe largest participation of olivine in the melting reac-tions in these melting ranges. The Fo content de-creases particularly rapidly close to the peridotitesolidus. Iron loss to the capsule after significantamounts of melt have formed may, however, partiallyaccount for the rapid change of Fo content.

Spinel. Spinel data from nodule 66SAL-l areplotted in Figure 5. Spinel is the most Cr-rich phasear-nong the minerals encountered in these experi-ments, containing up to 3.26 weight percent CrrO, at1375'C and 20 kbar . The Cr/ (Cr + Al ) of sp inel

1 4 0 00 1 0 2 0 3 0 L 0 5 0 6 0 7 0 8 0 9 0 1 0 0

Per cent me l t

Fig .3 Mel t ing curves o f nodu le 16 l l a t 20 and 35 kbar(anhydrous). Size of boxes as in Fig. l .


0 L l V l N E . 2 O k b o r

\ 'u' , ' , .r .n*,* rro

\ ' /$) ' -u.o,-- ,

30 40 50 60 70Pe r cen t me l t

100

Fig 4. Chemical t rends of o l iv ines as a funct ion of degree of mel t ing. Mineral assemblages descr ibe coexist ing phases in mel t ing

i ntervals.

9 7

96

9 5o

Is{,ot t+c'q?

\O)> 9 2

9 1

90802010

increases smoothly with increasing degree of melting.Similar observations were reported by Mysen andBoettcher (1975b) for spinels from 66SAL-l at about1000'C in the presence of HrO-saturated liquid. TheMg/(Mg * Fe) of spinel also increases smoothly withincreasing degree of melting (Fig. 5B). Like olivine(Fig. 4), spinel releases its Fe component most rap-idly during the initial stages of melting.

Orthopyroxene. Relevant chemical data for or-thopyroxene are shown in Figures 6 and 7. TheMgl(Mg * Fe; of this pyroxene as a function ofdegree of melting (Fie. 6,{) defines the phase fieldsmuch like that of olivine (Fig. a). Furthermore,Fe-Mg change during melting is most rapid in thebeginning of each melting range and with the fewestphases present. In contrast, alumina content does notfollow such a simple pattern (Fig. 68). Orthopyrox-ene of the aluminous sample 6654,L-l contains threeto four times more AlrOr in its melting range thandoes that of l6 l L The melt ing range of 66SAL-l isabout 100"C lower than that of l6 l l , probably con-tributing to this difference. The coexisting clinopyrox-ene may also be di f ferent in 66SAL-I and 16l l ,owing to the lower temperatures of the 66SAL-lmelting range (see below). Furthermore, the melting

behavior of orthopyroxene of l6 l l depends on thepresence of HrO. The orthopyroxene of l61l + 1.9weight percent HrO shows continuously-decreasingalumina content, whereas the AlrO, of orthopyroxeneof HzO-free l6l l remains constant within the twomelt ing ranges where i t is stable (Figs. l ,2, and 68).The chromium content of orthopyroxenes is gener-ally low (<0.5 weight percent CrrOr) and theirCr/(Cr + Al) define phase fields much like thosedefined by Mgl(Mg + Fe) (Fies. 6,{ and C).

The compositions of coexisting pyroxenes areplotted in the pyroxene quadrilateral in Figure 7. Thee n d m e m b e r s ( E n , F s , a n d W o ) a r e s i m p l yFe: Mg:Ca ratios. This procedure of calculation wasadopted because the Fe8+/Fe2+ of the pyroxenes arenot known and the amounts of various Tschermak'smolecules could not be calculated from the analyses.The effect of this simplification on the orthopyrox-enes is relatively small. Their position in the quad-rilateral is temperature-dependent. The amount ofmelt is unimportant as long as the clinopyroxene ispresent. However, the Wo component decreases rap-idly and the En component increases after the cli-nopyroxene is absorbed in the melt.

Clinopyroxe,rue. Relevant analytical data are plotted

g , p x . S O - L t

SPINEL(oosar - - r )

850 MYSEN AND KUSHIRO. COEXISTING PHASES OF PERIDOTITE

oox

+

IL

o

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0P e r c e n t m e l t

Fig. 5. Chemical trends of spinels from nodule 66SAL-1.Mineral assemblages describe coexisting phases in meltingintervals.

in Figures 7 and 8. Apparently, there is a crystal-chemical difference between the clinopyroxenes ofsample 665,4L-l on the one hand and those of l6l Ion the other (Fig. 7). The cl inopyroxenes from 1611(high-temperature melting interval) have only abouthalf the amount of Wo solid solutions that the665AL- I clinopyroxenes have (15-20 and 40-50 molepercent, respectively). Furthermore, the clinopyrox-

enes from l6l I are more iron-rich than those from665AL-1 (F ig.8A), despi te the h igher temperature offormation and the more magnesian coexisting crys-ta l l i ne phases o f l 61 l (F igs . 4 , 6 ,7 , and 8 ) . The Wocontent of clinopyroxene is simply ICa/(Ca * Mg +Fe)l X 100. However, about 10-14 percent of thetet rahedral cat ions in both 66SAL- l and l6 l1 c l i -nopyroxenes is occupied by a luminum. This Al Iv maybe subtracted as CaAlrSiO., leaving less than 5 per-cent wol lastoni te in l6 l I c l inopyroxene, or i t may beshared between (Mg, Fe, Ca, and Ca-Ti) Tscher-mak's molecules, yielding more than 5 mole percentWo. The net result is a Ca-clinopyroxene resemblingpigeoni te. The data of Kushiro and Yoder (1970) andHensen (1973) suggest between l0 and l5 mole per-cent wollastonite component in pigeonite in this pres-sure and temperature range. In that case, 50-75 per-cent of AIIV is present in these clinopyroxenes as(Mg,Fe) Tschermak's molecule. The Mg-Fe par t i -t ion ing between coexist ing pyroxenes of l6 l l -

lMg / (Me + Fe )1 .o " / [Mc / (Mc + Fe ) ] "o *I . I I - L l4-is in close agreement with the experimen-tal data on coexisting enstatitic and pigeonitic pyrox-enes by Hensen (1973) and with data on naturalcoexisting orthopyroxene and pigeonite by Kushiroand Nakamura (1970), further substant iat ing theidentification of the Ca-poor clinopyroxene froml6l l as pigeonite. Qual i ty analysis of Ca-poor cl i -nopyroxenes from 16l l + 1.9 weight percent HrOcould not be obtained because of significant inter-ference from quench pyroxene. Semi-quantitativecompositional data resemble those for clinopyrox-enes from experiments with anhydrous 1611.

The clinopyroxenes from 66SAL-l are diopsidic.The difference between these two samples is simplythe result of the temperature difference of their 20kbar melt ing intervals (1450"- l500oC and 1350'-1375'C respectively, for l6ll and 665AL-l). Kushiroand Yoder (1970) determined the lower stabilitylimit of iron-free pigeonite in the diopside-enstatitejoin to be 1450"C at 20 kbar, in close agreement withthe data shown in Figures 7 and 8.

The overall chemical trends (Fig. 8) are qual-itatively similar for Ca-rich pyroxenes from 665,{L-land l6l l . Both gain Mg and Cr and lose Wo and Jdcomponents (Figs, 8B and D) with increasing degreeof melting (Figs. 8A and C). Similar results wer.eobtained by Mysen and Boettcher (1975b) for diop-sidic pyroxene of 665,{L-l in the presence of HrO-saturated l iquid at about l000oC.

Liquid. Liquid compositions at20 kbar are shownin Table 3. The liquids can be classified into three

8 07 0

MYSEN AND KUSHIRO' COEXISTING PHASES OF PERIDOTITE 8 5 1

major groups (excluding the compositions within thefirst few percent of melting). The melt coexisting witholivine * 2 pyroxenes is olivine tholeiitic, followed bypicritic melt coexisting with olivine * orthopyroxeneand komatiitic (perioditic) melt coexisting with oli-v ine alone. Melts of both 66SAL-l and 161l at 12percent melt (within the dip of the melting curves;Figs. 1 and2) are nepheline-normative, with norma-tive nepheline contents decreasing from 7.6 percent at1.32 percent melt ( l6 l l ) to4.7 percent at1.77 percentmelt (Table 3). The phase assemblage does notchange as the liquid composition changes from neph-eline- to olivine- and hypersthene-normative athigher degrees of melting. The analytical data forcoexisting minerals, particularly the alkali content ofclinopyroxene (Fig. 8B), suggest that such read-justments of sol id solut ions are the major contr ib-uting factors in the formation of nepheline-normativeliquids in this low degree-of-melting range.

Variat ions of individual oxide contents of l iquidsplotted against degree of melting are shown in Figure9 for SiOr, Alro3, and CaO. In this figure, composi-tional changes over changes in phase assemblage aredrawn as straight vertical lines. Liquid compositionsmust change cont inuously, though rapidly, in suchregions. Because the nature of these changes is notknown, the composit ions are simply connected withstraight l ines in Figure 9.

Figure 9,{ shows how the si l ica content of 161ll iquids remains constant up to 45 percent melt , owingto the presence of pyroxene(s) of relatively constantsilica content. After pyroxene is lost, the liquids rap-idly lose SiOr, approaching komati i te composit ion.The temperature at 20 kbar is about 1550"C for 16l Iat this point. The presence of 1.9 weight percent HrOdepresses this si tuat ion to 1525'C and 60 percentmelting. This increase in degree of melting in thepresence of some HzO is to be expected, frop data onrelevant HrO-bearing systems (e.g., Kushiro et al . ,1968). I f 665,4L-l can be treated as possible sourcematerial for upper mantle magma, peridotitic koma-tiite may be produced at 85 percent melting at tem-peratures as low as l400oC.

The alumina contents of liquids are plotted in Fig-ure 9B and are controlled by the presence of highlyaluminous clinopyroxene (7-9 weight percent Alros).After the alumina-rich crystalline phase is melted, thealumina content of the l iquids drops rapidly and isessentially a function of the amount of crystals left inthe system.

The calcium contents of 161I l iquids (Fig. 9C) areinteresting because the Ca content of coexisting cli-

=

/

66 SAL_1

t 0 5 0 6 0 7 0P e r c e n t m e l t

Fig. 6. Chemical t rends of or thopyroxenes as a funct ion of

degree of melting. Mineral assemblages describe coexisting phases

in mel t ine intervals.

o

I

q

0 R T H 0 P Y R 0 X E N E ; 2 0 K b o r

\1611

66 SAL-1

66 SAL-1

1611* 1.9 w tahH2O

2 0 3 0 4 0 s 0 6 0 7 0 8 0

852

nopyroxene (pigeonite) falls rapidly within the melt-ing interval (Fig. 7), resulting in decreasing Ca con-tent of the coexist ing tholei i t ic l iquid. This si tuat ioncontrasts with that for 665,4L- l, where the coexistingcl inopyroxene is diopsidic and of relat ively constantCa content throughout the melting interval. The liq-uids of 6654L-l coexist ing with cl inopyroxene donot change their Ca contents significantly (Table 3).

Quality chemical analyses of liquids at 35 kbarwere unattainable owing to quenching problems.However, a few semiquantitative analyses inside theolivine + garnet * 2 pyroxene assemblage suggestalkali picrite composition. At successively highertemperatures, phase assemblages at 35 kbar are sim-ilar to those of 16ll at 20 kbar starting from itssolidus (Fig. 3), suggesting that the liquids in thesemelting ranges are similar as well. The 35 kbar melt-ing range of nodule l6l I would produce four distinctregions. The first 25 percent melt is alkali picritic (inthe presence of garnet), followed by olivine tholeiiticfrom 25 to 45 percent melting (in the presence ofolivine and two pyroxenes), followed by picritic from45 to about 75 percent melting (olivine * orthopyro-xene) and finally komatiite-type liquids in the high-est-temperature melting range, where olivine is theonly coexisting crystalline phase.


Fig. 7. Pyroxene quadr i lateral showing coexist ing pyroxenes of nodules l6 l I and 66SAL- l at 20 kbar.

2 0Fs---------)

1 0 2 0Fs

+

Discussion

Each melting range of peridotite, involving a spe-cific phase assemblage and a well-defined temper-Jture range, defines a specific melt composition thatvaries only within narrow compositional limits. Thewidth of these melting ranges depends on peridotitecomposit ions, inasmuch as the bulk composit ion ofperidotite is expressed by the proportions of peri-dotite minerals. In this sense, complex natural peri-dotite rocks behave similarly to model systems usedto describe melting behavior of peridotite in the up-per mantle.

The initial melting range (most closely resemblinginvariant) comprises 25 percent or more melt. Atpressures below the stability field of garnet, the liquidof the initial 25 percent or more melting resemblesolivine tholeiite composition. This liquid coexistswith olivine, 2 pyroxenes (orthopyroxene and pigeo-ni t ic c l inopyroxene in 1611;orthopyroxene and diop-sidic clinopyroxene in 665,4'L-l), and spinel. Fromneither sample can tholeiitic liquid be described by itsfour coexisting phases. Consequently, liquids of bothsamples have a reaction relation to their coexistingspinel peridotite residue. Reaction relations in vol-atile-free model peridotite were studied by Kushiro(1969) and by Kushiro and Yoder (1974) in the sys-

tem MgSiOr-CaSiOr-AlrOr. On the basis of thesestudies, two reaction relations can be deduced for thetwo nodules. The lower-temperature melting range of665,{L-l generates tholeiit ic l iquid by the reaction

Opx * Cpx * Sp = Ol * Liq,. (2)In the higher-temperature melting range of nodule| 6l l, the clinopyroxene is pigeonite, and the meltingreaction is

853

30 / . 0 50P e r c e n t m e l t

As would be expected from the data of Kushiro andYoder (1974), reaction 2 occurs at lower temperature(see Fig. l) and generates a l iquid (Liq') with aslightly higher Ca and Al content than that of Liq,(reaction 3) (Table 3). Unfortunately, l iquid compo-sitions could not be obtained from the 35 kbar gar-net-bear ing runs. This pressure of 35 kbar is wi th inthe range where the data on the simple CaSiOa-MgSiOr-AlzO3 system (Kushiro and Yoder, 1974)suggest the reaction

MYSEN AND KUSHIRO COEXISTING PHASES OF PERIDOTITE

CLIN0PYR0XENE; 20kbor

30 / . 0 50P e r c e n t m e l t

o

z

oz

o

oo

Fig. 8. Chemical t rends of c l inopyroxenes as a funct ion of degree of mel t ing. Mineral assemblages descr ibe coexist ing phases in mel t ing

i ntervals.

*##""t*

Ol * Cpx =r Opx * Liq, (3)

854 MYSEN AND KUSHIRO,, COEXISTING PHASES OF PERIDOTITE

Table 3. Composit ions of l iquids at 20 kbar

Star t ingM a t e r i a L :

T e m p . ( " C ) :

1 6 1 1

L450

1 6 1 I

L47 5

1 6 1 1 1 6 1 1

1500 7525

1 6 1 1 1 6 1 1 *

1 5 5 0 1 6 0 0

[ 6 I 1 *

I700

66 sAL- 1

1360

66 sAL- 1

t J t )

sio 2T102AL203Fe0' fMnOMcoCa0Na 20K20

Totals

4 3 . 8 0r . 40

1 2 . 7 01 3 . 7 00 . 3 0

1 2 . 0 01 3 , 7 0L . 4 00 . 6 0

99 .60

1 . 8 10

z o . t u7 , 5 7L . 4 0

a a 1 2

027 . 341 , 9 6

44.601 . 0 0

L 2 . 2 0L2 .500 . 20

1 4 . 3 0L3.201 . 1 00 . 5 0

99 .60

2 . 9 4L . 9 4

4 0 . t J

030.7 20

t . 3 9

50 . 200 . 8 0

1 2 , 2 09 . 6 00 . 3 0

13 . 80I 0 . 7 0r . 5 00 . 6 0

99 .70

1 1 ? ?

z + . ) z

00

22 .4018 . 981 6 . 1 81 . I 0

49 ,600 . 7 09 . l 09 , 7 00 . 20

1 9 . 6 09 . 2 00 . 900 . 5 0

q a q n

5 0 . 9 90 . 7 08 . 8 0

L 2 . 2 00 . t 0

15 .70o ?

0 . 399.40

48 .700 . 6 06 ; 1 0

1 7 . 8 00 . 1 0

18 . 507 . 2 00 . 7 00 . 3 0

100 . 00

46 ,300 . 4 04 ; 5 0

L4 .600 . 1 0

27 . 405 . 3 00 . 5 00 . 2 0

9 9 . 3 0

1 . 1 44 . 3 39 . 1 100

1 3 . 0 023.844 8 . 0 3

0 . 5 4

4 1 . 8 00 . 8 0

1 5 . 8 09 . 8 00 . 20

1 9 . 9 010 . 100 . 9 00 . 1 0

99 .40

0 . 4 01 . 7 0

3 7 . 6 01 1 )

08 . 5 20

46 .971 . 0 8

4 7 . 7 00 . 8 0

12.307 . 9 00 . 20

1 7 . 8 01 2 . 1 00 . 8 00 . 2 0

99 . 80

Cat ion ic Nom

oiAbAnNeL cc p xoPxo 1I 1

) a 9

7 . 8 91 8 . 8 6

00

2 0 . 5 72 5 . 6 6

0 . 9 s

r . 5 91 2 , 0 81 7 . 0 500

2 3 . O O

L 2 . 7 60 . 9 4

r . 7 76 . 2 6

1 2 . 5 800

L8 .4235 ,O7

0 . 8 3

0 . 9 86 . 8 3

2 6 - t J

00

23,8712.362 6 . I 9L . 0 2

SEar t ingMater la l :

T e m p . ( " c ) :

66sAL- 1

1400

L O I ] +

H2o*

1550

1 6 1 1 +H20

1500

1611+H20

1460

1611+H20

r47 5

1611+H20

7525

16 I1+H2o*

I6 00

1611+H20

17 50

s i02TIO2A 1 2 o 3F e o tMn0MgoC a ONa2OK20

Tota Is

4 6 . 8 00 . 9 0

t2.608 . 8 0o . 2 0

16 . 3012.60

L . 2 00 . 1 0

ffi

45 . r 01 . 1 0

1 1 . 0 01 3 . 6 00 . 3 0

13 .701 3 . 4 0

1 . 1 00 . 50

99 . 80

3 . 1 92 , 7 7

z ) . ) L

034.340

J U . ) O

t . 4 2

46 ,300 . 5 07 , 5 0

72 ,200 . 3 0

23.408 , 0 00 . 9 00 . 4 0

oo qn

2 . 4 68 . 0 7

1 4 . 5 100

19.078 . 2 4

46.960 ; 7 0

4 6 , 4 00 . 4 04 . 8 0

1 3 . 9 00 . 1 0

2 7 . 6 0s . 400 . 6 00 . 2 0

9e:Eo

l . J o

5 . 0 09 , 4 000

1 3 . 0 72 L . 6 748,960 . 7 0

47 ,O00 . 4 04 , 6 0

1 3 . 8 00 . 20

28 . 005 . 4 00 . 6 00 , 2 0

100 ; 20

4 6 . 1 00 . 4 04 . 4 0

1 3 . 7 0o . l 0

2 9 . 5 05 . 1 00 . 5 00 . 20

100 . 00

OrAbAnNe

c p xopxO II 1

0 . 5 8I U . ) O2 8 . 1 5

00

2 0 . ) J

L . 8 73 1 . 0 8

I . 2 3

1 . 3 04 . 8 78 . 8 700

, 1 r q

4 1 . 9 30 . 56

t . 2 3c . ) r8 . 5 100

l ? l R

1 8 , 9 153.94

n < ?

46 .70 48 .400 . 90 0 . 90

1 0 . 4 0 1 0 . 7 01 3 . 4 0 1 2 . 0 00 . 2 0 0 . 2 0

16 . 00 13 .309 . 9 0 1 2 , 7 01 . 5 0 1 . q 00 . 6 0 0 . 5 0

tt:6-b 1oo. oo

Cat ionic Nom

3 . 3 91 3 . 5 019 . 6300

23 .171 . 0 7

38. 04I . 2 0

2 . 8 9I t .452 I . 9 200

32 .68

23 .9L1 ) 1

*L iqu id cmpos i t ion ca lcu la ted f rm mass ba lance (p ropor t ions o f o l i v ine and coex is t lng l lqu id a re known) .'fTotal lron aa Feo.

Ga * Opx * Cpx a Ol * Liq (4)

to describe the initial melting of the natural peri-dotites.

At 20 kbar, about 25 percent melting of nodulel6l I is obtained at the invariant point described byreaction 3. As the temperature is increased further,the partial melt composition from nodule l6l I moves

down along an olivine-orthopyroxene liquidus sur-face and finally into the l iquidus volume of olivine.Simi lar reasoning appl ies to nodule 66SAL-1. Asmuch as 60 percent melt can be generated at theinvar iant point of nodule 66SAL-1, suggest ihg thatthis composition was itself generated at such iin in-variant point and therefore is not a parent but aproduct of partial melting in the upper mantle.

L lO UID horr)2 0 K b o r

P y r o x e n eo t r

a ,

Eo ,6 4=

9 0 1 0 0

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0P e r c e n t m e l t

Fig. 9. Chemical t rends of l iquids as a funct ion of degree of

melting. Mineral assemblages describe coexisting phases in meltirig

i ntervals.

855

Iri summary, the partial melts are compositionallycontrolled by their coexisting silicate minerals. In thespindl stability field, only three liquid compositionscan be obtained by partial melting of volatile-freeperidotite-olivine tholeiite, picrite, and peridotitickoniatiite. Within the stability field of garnet peri-dotite there are four types of liquids-alkali picrite,olivine tholeiite, picrite, and peridotitic komatiite.Clearly, a grid based simply on numeiical manipu-lation of proportions of minerals (as summarized byGreen, 1970) does not adequately describe the melt-ing behavior of volatile-free peridotite in the uppermantle.

AcknowledgmentsCri t ical reviews by Drs. F. R. Boyd, E. F. Osborn, R. V Dan-

chin, and H. S. Yoder, Jr . , are great ly appreciated.

References

Ar iderson, A. T. (1975) Some basal t ic and andesi t ic gases. Reu.

Ceophys , 13 ,37 -57 .Boettcher, A. L. , B. O. Mysen and P. J. Modreski (1975) Phase

relat ionships in natural synthet ic per idot i te-HrO and per i -

doiite-H,O-Co, systems at high pressure. Phys' Chem Earth,9,

955-968Bott inga, Y and D F. Wei l l (1972) The v iscosi ty of magmat ic

s i l icate l iquids: a model for calculat ion. Am. J Sci . ' 272,

438-47 5 .Bowen, N L. (1928) Euolution of the lgneous Rocks. Princeton

Univers i ty Press, Pr inceton, N. J.

Boyd, F. R. and J. L England (1960) Apparatus for phase equi l ib-

r ium measurements at pressures up to 50 k i lobars and temper-

atures up to 1750"C. J Geophyi . Res , 65, ' l4 l - '148.

Eggler , D H. (1974) i l f fect of CO, on the mel t ing of per idot i te.

Carnegie Inst Wash Year Book, 73' 215-224.- (1975) Pei idot i te-carbonat i te re lat ions in the system

MgO-SiO,-CO, Carnegie Inst Wash. Year 9ook,74,468-474'

Creen. D H. (1970) A review of exper imental evidence on the

origin of basaltic and nephelinitic magmas. Phys Earth Planet.

l i ler iors. 3.221-235.- sni A. E. Rin$wood ( 1967) The genesis of basal t ic magmas.

Contrib Mineral. Petol , /J, 103-190.

I rv ing, A J. and P. J. Wyl l ie (1975) Subsol idus and mel t ing re la-

t ionshipi fo i calc i te, magnesi te and the jo in CaCOr-MgCO. to

36 kb Geochim. Cosmochim. Acta, 39,35-53-

Hensen, B. ,J. (1973) Pyroxenes and garnets as geothermometers

and geobarometers. Carnegie Inst' Wash. Yeat Book, 72'

527-534.Jackson, E. D. dnd T. L. Wright (1970) Xenol i ths in the Honolulu

volcanic series, Hawaii J. Petol , I I, 405-430'

Kushiro, I . (1969) The system forster i te-diopside-s i l ica wi th and

without water at h igh pressures. Am. J Sci , 267'A,269-294.- (1972) Effect of water on the composition of magmas

formed at h igh pressures. J. Petrc l , l3 ,3 l l -334.- (1973) Part ia l mel t ing of garnetlherzol i te f rom kimber l i te

at high pressures In P. H. Nixon, Ed', Lesotho Kimberlites, p'

294-300. Lesotho National Development Corp , Maseru'- nn6l Y Nakamura (1970) Equi l ibr ium relat ions oI hypers-

thene, pigeonite and augite in crystallizing magmas: microprobe


o

c@

@o

co

o

=

o

co

@

o

Eoo=

C l i n o p y r o x e n ep r e s e n t

N o C l i n o -

P y r o x e n e

N o p y r o x e n e

C l i n o p y r o x e n eP r e s e n t

N o - C l i n o p y r o x e n e

N o p y r o x e n e

856

study of a pigeonite andesite from Weiselberg, Germany. AmM ineral.. 55. 1999-2016.

- and H S. Yoder, Jr. (1970) Stabi l i ty of iron-free pigeonitein the system MgSiOr-CaMgSirO". Chrnegie Inst. ll/ash. YearBook, 68,226-229.

- and - (19j4) Formation of eclogite from garnetlherzol i te: l iquidus relat ions in a port ion of the systemMgSiOr-CaSiOe-Al,Oe at high pressure. Carnegie Inst. l4tash.Year Book,73,266-269.

and B. O. Mysen (1976) Viscosity of basalt andandesite magmas at high pressures. EOS, 57, 354.

and M. Nishikawa (1968) Effect of water on themelt ing of enstat i te. Geol. Soc Am. Bull . ,79, 1685-1692.

Merri l l , R. 8., J. K. Robertson and P. J. Wyll ie (1970) Melt ingreactions in the system NaAISLOT-KAlSisOs-SiOr-HrO to 20kilobars compared to results from other feldspar-quartz-HrOand rock-HrO systems. J. Geol.,78, 558-570.

Mysen, B O., R. J. Arculus and D. H. Eggler (1975) Solubi l i ty ofcarbon dioxide in natural nephelinite, tholei i te and andesitemelts to 30 kbar pressure. Contr ib. Mineral Petrol. ,53,22' l-239.

- and A. L. Boettcher (1975a) Melt ing of a hydrous mantle. I .Phase relat ions of natural peridoti te at high pressures and tem-peratures with controlled activities of water, carbon dioxide andhydrogen. J. Petrol. , 16,520-548.

- and - (1975b) Melt ing of a hydrous mantle. I I . Geo-chemistry of crystals and l iquids formed by anatexis of mantle


peridotite at high pressures and temperatures as a function of

controlled activities of water, hydrogen and carbon dioxide. "/.Petro l . . 16.549-592.

- , D. H. Eggler , M. G. Sei tz and J. R. Hol loway (1976)

Carbon dioxide in s i l icate mel ts and crystals. I . Solubi l i ty mea-

surements. Am. J. Sci . , 276,455-4'19.- and M. G. Sei tz (1975) Trace element part i t ioning deter-

mrned by beta-track mapping: an experimental study using car-

bon and samarium as examples. " / . Geophys. Res ,80,2627-2635.Nixon, P. H and F R. Boyd (1973) Petrogenesis of the granular

and sheared ul t rabasic nodule sui te in k imber l i tes. In P. H.

Nixon, Ed., Lesotho Kimberlites, p. 48-56. Lesotho National

Development Corp. , Maseru.

Shaw, H. R. and E. D. Jackson (1973) L inear is land chains in the

Pacific: results of thermal plumes or gravitational anchors? "/.Geophys Res., 78, 8634-8652

Shimizu, N. (1974) Rare earth e lements (REE) in garnets andcl inopyroxenes f rom garnet lherzol i te nodules in k imber l i te.

Carnegie Inst. Wash. Year Book,73,954-961Waff , H. S. I t975) Pressure- induced coordinat ion changes in mag-

matic liqirids. Geophys. Res. Lett , 2, 193-196.

Yoder, H S, fr. ltlZe; Generution of Basaltic Magma. National

Academy of Sciences, Washington, D. C.

Manuscript receiued, Oclober 7, 1976: accepled

for publication April i,9, 1977.

compositional variations of .tt?f;yil1ff?;11,[ degree of ... · american mineralogist, volume 62,...

Documents